npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@instruments/visual-grounding

v0.1.0

Published

Stream visual regions from detection through semantic enrichment and optional segmentation.

Readme

@instruments/visual-grounding

One streaming contract for evidence-bearing observations of visual surfaces.

import { analyzeVisualSurfaces } from "@instruments/visual-grounding";
import { createOpenRouterDetector } from "@instruments/visual-grounding/openrouter";

const pipeline = {
  detector: createOpenRouterDetector({ model: "openai/gpt-5.6-luna" }),
};

for await (const event of analyzeVisualSurfaces(pipeline, {
  image,
  instruction: "Detect every distinct interior material surface.",
  capabilities: ["geometry", "semantics", "segmentation"],
  sessionId,
})) {
  if (event.type === "surface.detected") {
    drawBox(event.surface);
  }
  if (event.type === "surface.refined") {
    inspect(event.capability, event.observation);
  }
}

Capabilities

Consumers request outcomes rather than pipeline stages:

  • geometry detects surface occurrences and normalized boxes;
  • semantics adds descriptions and controlled vocabulary fields;
  • segmentation produces a surface mask;
  • ownership identifies the eroded, source-sized mask interior whose pixels are eligible for color measurement;
  • color returns the observed palette, spot, ColorVec-compatible embedding, exact pixel counts, ownership mask, and source dimensions.

Requesting ownership or color plans segmentation automatically. Requesting color also plans ownership. Ownership completes only when a color profiler returns its exact eroded data-url mask; the raw segmentation mask is not reported as measured pixels.

Events and outcomes

surface.detected streams usable geometry immediately. Every surface.refined event identifies the capability that settled and carries a complete immutable observation snapshot. complete contains both the legacy surfaces collection and the evidence-bearing observations collection.

Each requested capability settles as one readonly outcome:

  • completed carries a value and a receipt naming the method;
  • failed carries a technical error message;
  • abstained carries the evidence-based reason the package could not make the observation.

A failed or abstained refinement does not discard successful geometry or other completed capabilities. Segmentation failure causally abstains ownership and color, while a profiler exception fails ownership and color without discarding segmentation.

Legacy detail presets

detail remains as a compatibility preset for existing callers: "boxes" maps to geometry and "masks" maps to geometry plus segmentation. When both fields are supplied, capabilities takes precedence. Detail callers continue to receive surface.enriched, surface.segmented, and surface.failed; capability callers receive surface.detected, surface.refined, and complete.

Provider and model choices remain behind adapters. Catalogue retrieval, ranking, and product truth remain downstream.